projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 392e16c) | patch
[AsyncToLLVM] Align frames to 64 bytes
authorBenjamin Kramer <benny.kra@googlemail.com>
Mon, 17 Jan 2022 17:23:44 +0000 (18:23 +0100)
committerBenjamin Kramer <benny.kra@googlemail.com>
Mon, 17 Jan 2022 17:51:42 +0000 (18:51 +0100)
commit5acd6e05221574643feb8477ad07e89fd766ec53
tree1e352f52427db78891aaaf78a2590a8cb17e3f78tree | snapshot
parent392e16c27ffce3c88d6d14050327bb7409756900commit | diff
[AsyncToLLVM] Align frames to 64 bytes

Coroutine lowering always takes the natural alignment when spilling to
the frame (issue #53148) so using AVX2 or AVX512 in a coroutine doesn't
work. Always overalign to 64 bytes to avoid this issue until we have a
better solution.

Differential Revision: https://reviews.llvm.org/D117501
mlir/lib/Conversion/AsyncToLLVM/AsyncToLLVM.cpp diff | blob | history
mlir/test/Conversion/AsyncToLLVM/convert-coro-to-llvm.mlir diff | blob | history
Domain: System / Toolchain;